Using SOM and Data Mining Models to Establish an Intelligent Web Mining System

• Main Authors: Yao-Bai Fang, 方耀白
• Other Authors: Yo-Ping Huang
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/66942706442048347636

碩士 === 大同大學 === 資訊工程研究所 === 89 === Electronic commerce rises sharply because of the growth of IT (Information Technology) and Internet. Customer oriented product and selling become a major tendency of commercial affairs little by little. In electronic commerce many behaviors of commerce can be done by a website. But the behavior model of electronic commerce is more passive than the behavior module of traditional commerce. If we can apply an intelligent system to web then it can increase active property to thrive trade. WWW is an important information medium on Internet which can offer much useful information. Web sites also include all kinds of information. How to provide a convenient user interface to various users and how the web site is used by user are all very important to system administrator. Traditional website does not give CRM (Customer Relationship Management) service to users. This results in user can not fast find information that he/she wants to get and is interested. In accordance with various users we want to use a technique of neural network to automatically get user cluster and combine the technology of data mining to analyze different users’ glance customs and users access patterns. Analyzing by such a technology we can get the data of interests from any kind of users or the characteristic rules of using websites. The result can dynamically offer users to view related pages on the website by ASP or PHP, which is the website program processing on server that can provide users about their interests content and products. Further, we can actively offer users the service of CRM, or we can recommend users their interesting products to create business opportunities and to let them find what they want faster. In this thesis, we present a method to filter log file and get the believable data to be training data in our system. In a large website can have many kinds of users, first we apply SOM network to cluster users, then analyze ever user cluster’s path pattern. In the last step, we add a tendency concept to the original algorithm for mining user path patterns.